Coolscan V ED first impressions/questions

[mszargar]

Well, I cracked and bought one, as I need to practice with film for a while. The scanner is clean and seems to be from a later batch (4xxxxx serial number).

First impression: It is the first time I see a scanner can get the colors right out of the box. I had tried Canon and Epson flatbeds before as well as 35mm Plusteks with some frustration. This one works.

Downside: I don't know if this is my lack of experience or what, but I find the sensor very noisy.

Velvia scans correctly, although the details in denser parts are replaced with deep blacks. ME in Vuescan or DDE in Nikon Scan do not help to extract more details.

But except from Velvia (meaning slide film) I get too much color noise on anything else (meaning any negative film). I have tried Portra 800, Portra 400, Konica VX200, and Fuji ProPlusII with mediocre results. Finding the right color correction mechanism for each of these films is already a hassle, but even when I find it, there is just too much noise in form of bright color points (very bright blue or red, or green...). Some of the results are absolutely worse than those from Noritsu labscans. I have also tried multi-sampling to no avail.

Interestingly, the linear RAW TIFF from Vuescan seems to be less noisy than the positive I get from either Vuescan or Nikon Scan. The details are also much more present in dynamic range extremities.

Last but not least, the scanner light is just too hard on the film, and it reveals too much the grains and imperfections (scratches, dust). I get rid of imperfections easily in Photoshop (for b&w - infrared cleans color scans perfectly), but the hard grains, that I guess are combined with a good deal of noise, make my B&Ws just too messy. This is true even for the modern Delta 100 and T-Max 100 Pro.

I appreciate any experience you share with me. I will try to upload some samples soon. A link to a personal scanning workflow guide will be much appreciated, specially if it contains a section with film color correction and negative color cast removal using lab color.

 

[johnny.moped]

The Coolscans lightsource is really terrible (and that was the main reason why I got rid of it).
I did some comparisons with a Flextight X1 a while ago.
And even the Flextight has kind of a hard lightsource (the X5 has a much softer light).

Click to view in full res.

 

[Michalm]

Noise visible on the screen did never show up on any of the prints i made from V scans , unless incorrect contrast curve or too much micro-contrast sharpening is applied to the picture . I would ignore grain issue.

 

[biomed]

I have had a Coolscan V ED for a number of years. I bought it to replace my dead Canon 4000US that died. I have had good scans without excessive noise. When I upgraded to a 64 bit OS, I kept a dedicated computer running 32 bit XP for the scanner. The files were transferred to my photo editing computer via hot swappable HDs. This allowed me to be editing photos while scanning negatives.

I had my best results using the Nikon software. For some reason Vuescan would never seem to work for me. I downloaded the Win 7 version of Vuescan, but haven't used it yet. I also have a Super Coolscan 4000 ED from an equipment trade. I hope to get up and running after my upcoming retirement. BTW, there is an excellent book on Vuescan called the Vuescan Bible.

 

[mszargar]

Thanks for all the feedback.

I understand that the LED light in V ED is rather hard. Yet, I think the nature of noise that I get is quite different from what I see in Johnny's post. I get really huge, ugly clumps of noise in very bright colors that do show up in prints. This time I attach a sample. The noise looks like clipping indicators in photo editing software!

The original photo:

100% crop @4000 dpi:

Any suggestion?

Btw, I tried scanning Ektar, and it comes out perfectly. The photo above was taken using Portra 800. No sharpening, automatic color restoration (Nikon Scan) and a very slight bit of grain reduction (Again, in Nikon Scan).

 

[brbo]

In Vuescan you can lock exposure so that bright parts of the negative (shadows) won't clip. You can also use multisampling in Vuescan to reduce the noise.

I don't know if/how you do that in Nikon Scan.

 

[mszargar]

Nikon Scan has an option called DDE that IMHO does the same job as multiexposure... But I am not sure if these blue/green dots are exactly color clips. For one thing they are visible in VueScan as well. Plus, one thing that is missing in VueScan is a powerful color correction engine. For instance, this photo, shot on Portra 800 under Tungsten light can only be color corrected in Nikon Scan. I have tried SilverFast, ColorPerfect, VueScan and Photoshop Lab mode curves to no avail. Still, if I can make sure these are clips, I will try to fix the colors otherwise. One clue: when I open this files in Adobe Lightroom the noise is horribly accentuated. Can it be that the color mode is not being correctly assigned or something like that?

 

[aldobonnard]

Nikon Scan has an option called DDE that IMHO does the same job as multiexposure...

Actually it does not. it's only software based, somewhat useless.
Vuescan with multi exposure and multi-sampling gives good results.

The Coolscan V is a decent scanner, fast. But its dynamic range is nowhere near as good as the Minolta Elite 5400 and 5400 II.
I am entirely sure of this, because owning a 5000ED, a DI 5400, a DI 5400 MkII, I have compared all three.
The Nikon are all right for speed and negative films with odd colours (using a tiny bit of ROC correction).
The KM DI 5400 I and II are great for slides and has a 100 times better ICE dust removal than the Nikon 5000ED, but the 5400 II ROC implementation is not great.
In terms of noise, the Nikon 5000ED produces poor noise patterns, big blotches. The Minolta produce fine grain instead, despite its true resolution of 4700-4800dpi (Nikon Coolscan 5000ED gives only a true optical of 3600dpi). Multisampling is necessary in both scanners to avoid noise banding.
As concerns B&W, I have found that Vuescan gives the best result but your scans have to be retouched in Lightroom afterwards.
The Coolscan gives excellent good results in B&W with Vuescan, albeit grainy scans, yes. But... once you print the scans, the grain is not really visible, so don't worry.
The problem is the shape of Nikon Scan's noise grain patterns, it is awful...

 

[philipus]

I've used the V for many years and, yes, it does have a fairly harsh light source. Still it can be used with incredibly good - money for quality - scans. I do disagree slightly that it has poor dynamic range. At least for my purposes I have been very happy with it. With film there's so much information in the negs that one can usually get stuff from seemingly blown highlights too. I'm also happy with how it treats shadows, but of course ymmv etc.

I used NikonScan extensively until I replaced it with Vuescan as I moved to Mac and I am happier with the results from Vuescan. That being said, NikonScan tended to give better results "out of the box" - Vuescan requires being set up to give the best results.

As for the colour noise, there are a few ways to deal with it. One is to run the noise reduction filter in Photoshop (or ACR if you choose to open the TIFF/JPG there). This is pretty effective in desaturating the noise to make it blend in with the grain. But it is something I fear one has to accept for some images - as always, a properly exposed neg will scan better than one that is not well exposed.

Pepper grain is something that will also appear with the V (again depending on the image of course). In my experience it is less likely to appear on slides than colour negs. I personally have not found it annoying but I know there's a pretty big discussion about it in various fora (including photo.net where there were threads a few years ago abou this).

I don't make colour corrections in Vuescan but leave that to ColorPerfect which is what I use to "develop" my images. My aim when scanning is to get the flattest looking image I can, thus preserving the most data in the image. This usually means I scan with the preset "None" under the Color tab in Vuescan (I do usually have a film profile selected, but depending on the image the result may be better using the generic film type - for both negs and slides).

CP works very well, I find. I know how to colour correct also in Photoshop but CP works best for me and is a bit more accessible, I find, because it has all the various controls in one window. I also find that CP will give me some control over how color noise appears, actually.

LAB is obviously great for colour corrections and as necessary I will use this for extra colour corrections if I'm unhappy with the results in CP.

 

[mszargar]

Well, apparently I do get less noise in Vuescan. I don't know how to color correct Portra 400 and 800 in Vuescan/ColorPerfect though. Specially tungsten light seems to leave an uncorrectable effect. One touch of ROC in Nikon Scan changes everything, though. But then it leaves me with these horrible huge noise spots. The noise I get with ColorPerfect is fine enough to remove with denoise, but the one from Nikon Scan can not be removed without sacrificing a whole lot of details in the scan.

I see what aldo means by limited dynamic range. I do agree that I do my best scans on V ED when I lock the exposure. A broader dynamic range could be more tolerant to over/under exposures during scan, which in turn means spending less time tweaking scan preferences...

 

[johnny.moped]

Tungsten light might be the problem.
Nikon Scan may let you correct that, but at the same time it has to boost one of the color channels in such a extreme way that you get this kind of noise.
Are you familiar with color balancing in Colorperfect by clicking on a grey area?
Try to hold down the shift-key while doing this and it also lets you do extreme color balancing.

 

[mszargar]

Johnny, thanks for the suggestion, I will try it tomorrow.

For now the problem is not color correction, but the blue noise. I am starting to doubt about this scanner!

I have been learning exposure and gain control during the last few days to make sure I can squeeze out the maximum of information from my negatives. I have found out that this blue noise (and yes, it is nearly always blue) is not color clipping!

Consider the following example followed by two 100% crops:

See the blue noise in the dark hair.

A completely transparent part of this film (ProPlusII) scans as black with no clipping, all color curves (RGBI) centered around a single axis with the following values:

RGB Exposure: 2.2
IR Exposure: 1.03
Red Gain: 0.93
Green Gain: 0.94
Blue Gain: 1

Now, the above photo did not clip up to 3.2 RGB Exposure. I scanned it at 3.2, 3, 2.6, and 2.2 with and without multi-exposure and the result is still the same: blue noise in the hair. At 2.2 the whole color distribution covered only a fraction of the available gamut, leaving no chance to clipping. And as you see, this time the photo has been taken in daylight and absolutely no color correction has been done. The photo you see is just the flat file out of Vuescan with a slight s-curve applied in Photoshop. The blue noise exists even in Vuescan's raw file!

Any idea what this can be?

 

[brbo]

Since your slide film scans come out OK I don't think there is a problem with a scanner.

Based on my (admittedly pretty limited) understanding of RGB color model principles and on my own experience with scanning, the problem of 'blue noise' is one directly related to C-41 orange color mask.

Scan small part of an unexposed part of C-41 film as slide film. If you use Photoshop to get a uniform color (getting rid of the noise pattern/deviance) of the scanned image you will see that the value of the orange mask for the R channel is the highest, followed by G channel and B channel as the lowest. For this reason the scanner (when you set the media type to color negative in Vuescan) will use higher (longer) exposure when reading G channel and even higher when reading B channel. Now, the scanner/Vuescan might not figure out the correct individual exposures for individual channel. So, you have to help them.

I'm not at home to give you exact step by step in Vuescan (I can do it later if you want).

The basic idea is to:
1. get RGB exposure lock
2. determine the correct individual RGB gains

Nr.1 is easy and you've already done it. Nr.2 is a bit trickier. You will need to switch your input media type to slide film. On color tab set color balance to none, do preview and determine how good your individual RBG exposures cancel out the orange mask. Repeat/readjust individual RGB gains until you are satisfied. You should get a neutral light color in preview (I aim for the RGB values around 210). Graph and RGB tooltip values will help you determine how close you are.

Also, for lowest noise mutisampling is essential.

I have Minolta 5400 v1 and I can get a bit better shadow noise (less blue noise) with carefully adjusted individual RGB exposures, but I rarely do this. I scan to raw and invert with ColorPerfect where I don't use any color film profiles (I use a preset for basic B/W) and only adjust neutral point and 'black' and 'white' sliders to get zero clipped pixels. I then adjust contrast/shadows/highlights in Lightroom. In 95% I don't even touch individual curves or use any color correction on the files that I get from ColorPerfect.

As for your posted scan and Vuescan settings... I learned that Vuescan does a pretty poor job at inverting color negatives (ColorPerfect is really good and the (two?) people behind it know what they are doing - they also publish a lot of very useful articles on scanning). Also, your individual RGB gains are suspiciously close. I usually arrive at the values that are far more apart. Almost 1 'stop' of difference between R, G and B (e.g. R 0.5, G 1.0, B 2.0).

Still, shadows on color negative film will be noisy (especially when looked at on monitor screen at 100%) if you shoot at box speed and/or don't bias your exposure heavily for the shadows.

 

[brbo]

I looked up the specs of your scanner. It seems to have 14bit A/D converter. My Minolta 5400 seems to have 16bit. That could explain why I'm seeing less blue noise (but blue noise is still more noticable) and that it doesn't make that much of a difference for my scans if I set correct individual RGB exposures. I guess the 'baked-in' individual exposure correction for negative film is close enough and vast bitspace prevents excessive clipping.

I think that you should be able to reduce color noise in your Nikon V scans if you can get the individual exposures matched perfectly to particular orange mask.

 

[kanzlr]

That looks pretty much exactly like what I got out of my Coolscan 4000.